Write a code based on ICP algorithm to fully reconstruct object from depth sensor

There is a video sequence from Intel RealSense Depth camera in format of .bag file (can be converted to the PLY files sequence). Video has been captured in 360 degree around human head and shoulders. The task is to prepare code (C++ or similar) based on opens-soursed algorithm ICP (as a part of PCL library) to fully reconstruct human head with high quality. During coding you should clear source data preliminary from noises and filter untypical points. Clearing/Filtering process should be fully automated, no any manual pretreatment. For Clearing/Filtering process you can also use existing algorithms which you familiar with. Clearing/Filtering process is mandatory stage to solder individual scenes in better way. It's also allowed to use alternative existing algorithms or ICP modifications to reach the best result if you belive that classical ICP algorithm is not ideal code for this task. Reconstructed object (human head) shold has tolerance no more 1 cm from real object in any dimension (referenced dimensions). Referensed dimensions will be provide to you later. Resulting code should by fully autometed and doesn't require any pre-treatments. Resulting code should be comiled into exe file wich has input as a .bag file from Intel RealSense depth camera. On output should be fully recunstructed head in the fromat of .ply file. Program should works with any other .bag samples from RealSense camera with the similar objects. You can use downsampling in algorithm to achieve better results and to decrease time to execution. Program should works on x86-based systems. It's allowed to use CPU cores in parallel or additionaly involve GPU to increase program performance. Please keep in mind that during the work you can request any additional video sequenses in .bag file format with different capturing conditions (for instance - distance to object, speed of capturing, voxel resolution etc.)